A real-time scheduling method for a water conservancy project system based on neural network technology

Abstract

The present invention relates to the technical field of flood control management, and specifically to a real-time scheduling method for a water conservancy project system based on neural network technology, including: obtaining real-time data of the water conservancy project system at fixed time intervals; using a CNN model to perform image recognition on satellite cloud images to obtain a rain cloud area map and a rain cloud area edge map; calculating the edge roughness coefficient, internal density coefficient, and rain cloud thickness coefficient of the rain cloud area map and the rain cloud area edge map to obtain a rainfall coefficient; using the optical flow method to analyze the rain cloud area maps at consecutive time intervals to obtain the moving speed, direction, and shortest distance from the rain cloud to the water conservancy project area to obtain a rain cloud influence coefficient; using a deep learning model to predict the future water level of the reservoir and calculating the vulnerability coefficient based on flood control data; calculating the flood control coefficient based on the rainfall coefficient, rain cloud influence coefficient, and vulnerability coefficient, and constructing a flood control regulation model, effectively improving the flexibility of the real-time scheduling of the water conservancy project system.

Description

Technical Field

[0001] The present invention relates to the technical field of flood control management, and particularly to a real-time scheduling method for a water conservancy project system based on neural network technology. Background Art

[0002] Traditional water conservancy project scheduling methods usually rely on experience and historical data, and it is difficult to adapt to the real-time changing environment. This results in insufficient timeliness in responding during floods, especially in extreme weather conditions, where sudden rainfall cannot be quickly responded to, thus affecting the reservoir's water storage and flood discharge decisions and unable to effectively ensure the safety and water supply stability of the reservoir. To improve the scheduling efficiency and flexibility of water conservancy projects, more and more research has begun to focus on real-time scheduling methods, which have improved the scheduling response ability. However, existing methods often lack a comprehensive analysis of rain clouds, terrain factors, and social factors, and also lack the use of this information for effective reservoir management, thus limiting the dynamics and accuracy of scheduling strategies. Especially when facing complex meteorological changes and multiple scheduling constraints, the flexibility needs to be further improved.

[0003] Therefore, a real-time scheduling method for a water conservancy project system based on neural network technology is proposed. Summary of the Invention

[0004] The purpose of the present invention is to provide a real-time scheduling method for a water conservancy project system based on neural network technology, which obtains the real-time data of the water conservancy project system at fixed time intervals; uses a CNN model to perform image recognition on satellite cloud images to obtain a rain cloud area map and a rain cloud area edge map; calculates the edge roughness coefficient, internal density coefficient, and rain cloud thickness coefficient of the rain cloud area map and the rain cloud area edge map to obtain a rainfall coefficient; uses the optical flow method to analyze the rain cloud area map of consecutive time intervals to obtain the moving speed, direction of the rain cloud, and the shortest distance between the rain cloud and the water conservancy project area to obtain a rain cloud influence coefficient; uses a deep learning model to predict the future water level of the reservoir and calculates the vulnerability coefficient according to flood control data; calculates the flood control coefficient according to the rainfall coefficient, rain cloud influence coefficient, and vulnerability coefficient, and constructs a regulation model, effectively improving the flexibility of the real-time scheduling of the water conservancy project system.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A real-time scheduling method for a water conservancy project system based on neural network technology, comprising:

[0007] Obtaining the real-time data of the water conservancy project system at fixed time intervals, the real-time data including satellite cloud images, rain cloud vertical profile data, and flood control data, and preprocessing the satellite cloud images to obtain a first image dataset;

[0008] Use a CNN model to perform image recognition on the first image dataset. The CNN model is used to extract the spatial features of rain clouds to obtain a rain cloud area map and a rain cloud area edge map. The image recognition process is as follows: perform feature extraction based on the first image dataset to obtain low-level image features; form a first branch network and a second branch network according to the low-level image features. The first branch network is used to segment the rain cloud area and output the rain cloud area map, and the second branch network is used to extract the edge of the rain cloud area and output the rain cloud area edge map.

[0009] Calculate the edge roughness coefficient and the internal density coefficient based on the rain cloud area map and the rain cloud area edge map; calculate the rain cloud thickness coefficient based on the rain cloud vertical profile data, and calculate the rainfall coefficient based on the edge roughness coefficient, the internal density coefficient, and the rain cloud thickness coefficient.

[0010] Use the optical flow method to calculate the rain cloud movement speed and the rain cloud movement direction for the rain cloud area map at continuous time intervals, and calculate the shortest distance between the central rain cloud and the water conservancy project area within the rain cloud area map; calculate the rain cloud influence coefficient based on the rain cloud movement speed, the rain cloud movement direction, and the shortest distance.

[0011] Use a deep learning model to predict the future water level of the reservoir based on the flood control data, and calculate the vulnerability coefficient based on the flood control data and the future water level of the reservoir.

[0012] Calculate the flood control coefficient based on the rainfall coefficient, the rain cloud influence coefficient, and the vulnerability coefficient, and establish a flood control regulation model based on the flood control coefficient. The flood control regulation model is used to perform real-time regulation on the water conservancy project system.

[0013] Further, the flood control data includes: historical flood defense data, reservoir historical data, reservoir real-time data, terrain and geographical data, building density, population density, and hydrological facilities.

[0014] Further, the preprocessing includes:

[0015] Perform color space conversion on the satellite cloud image to obtain a first cloud image;

[0016] Correct the illumination and radiation of the first cloud image to obtain a second cloud image;

[0017] Determine the downsampling ratio according to the task requirements, and perform downsampling on the second cloud image according to the downsampling ratio using bilinear interpolation to obtain a third cloud image;

[0018] Remove noise from the third cloud image using Gaussian filtering to obtain the first image dataset.

[0019] Furthermore, the training steps of the CNN model include:

[0020] Step S1: Obtain historical satellite cloud images and perform the preprocessing to obtain a historical cloud image dataset;

[0021] Step S2: Label image tags for the historical cloud image dataset to obtain a second image dataset, and input the second image dataset into the CNN model; wherein, the image tags include rain cloud regions, non-rain cloud regions, edge points, and non-edge points;

[0022] Step S3: Perform a convolution operation on the second image dataset using a convolutional layer to output a first feature map;

[0023] Step S4: Perform a pooling operation on the first feature map using a ReLU activation function and a pooling layer to output a second feature map;

[0024] Step S5: Perform a segmentation operation on the second feature map using the first branch network to output the rain cloud region map;

[0025] Step S6: Perform edge feature detection on the second feature map using the second branch network to output the rain cloud region edge map;

[0026] Step S7: Calculate the segmentation loss of the first branch network using a cross-entropy loss function, calculate the edge detection loss of the second branch network using an IoU Loss function, and perform a weighted sum of the segmentation loss and the edge detection loss to obtain a combined loss value;

[0027] Step S8: Update the CNN model according to the combined loss value;

[0028] Step S9: If the number of iterations reaches a preset number, stop the training; otherwise, repeat Steps S3 to S9.

[0029] Furthermore, the calculation process of the rainfall coefficient includes:

[0030] Use the box-counting method to calculate the fractal dimension of the edge of the rain cloud region edge map to obtain the edge roughness coefficient;

[0031] Calculate the variance of the gray values of the pixel points of the rain cloud region map to obtain the internal density coefficient;

[0032] Use the rain cloud vertical profile data to calculate the vertical height difference between the cloud top and the cloud bottom, and calculate the thickness change rate of the vertical height difference for consecutive time intervals to obtain the rain cloud thickness coefficient;

[0033] Perform a weighted sum of the edge roughness coefficient, the internal density coefficient, and the nimbus cloud thickness coefficient to obtain the rainfall coefficient.

[0034] Further, the calculation process of the nimbus cloud influence coefficient includes:

[0035] Use the optical flow method to calculate the motion vector field of each pixel point for the nimbus cloud region map of consecutive time intervals. The direction of the motion vector field is the direction of nimbus cloud movement, and the average modulus length of the motion vector field is the nimbus cloud movement speed.

[0036] According to the nimbus cloud coordinates and the water conservancy project area coordinates, use the Euclidean formula to calculate the obtained shortest distance.

[0037] Calculate the cosine value of the angle between the nimbus cloud movement direction and the direction from the central nimbus cloud to the water conservancy project area, and multiply the cosine value by the nimbus cloud movement speed to obtain the second nimbus cloud movement speed.

[0038] Divide the shortest distance by the second nimbus cloud movement speed to calculate the obtained nimbus cloud influence coefficient.

[0039] Further, the calculation process of the vulnerability coefficient includes:

[0040] Input the reservoir historical data into the basic model for training to obtain a trained deep learning model; wherein, the reservoir historical data includes historical water level data, meteorological data, and reservoir inflow and outflow.

[0041] Input the reservoir real-time data into the trained deep learning model for prediction to obtain the future water level of the reservoir.

[0042] Calculate the ratio of the future water level of the reservoir to the reservoir storage capacity to obtain the reservoir storage coefficient.

[0043] Obtain the flood occurrence times and flood losses according to the flood defense historical data to get the flood occurrence frequency and flood loss rate, and multiply the flood occurrence frequency by the flood loss rate to obtain the flood risk coefficient.

[0044] Obtain the slope and soil type according to the terrain and geographical data, and divide the grades according to the slope and the soil type to obtain the terrain score; calculate the average terrain score of the water conservancy project area according to the terrain score to obtain the terrain vulnerability coefficient.

[0045] Perform a weighted sum of the building density and the population density to obtain the social vulnerability coefficient.

[0046] Perform a weighted sum of the reservoir storage coefficient, the flood risk coefficient, the terrain vulnerability coefficient, and the social vulnerability coefficient to obtain the vulnerability coefficient.

[0047] Further, the calculation of the flood control coefficient includes:

[0048] Normalize the rainfall coefficient, the rain cloud influence coefficient, and the vulnerability coefficient respectively, and perform weighted summation on the normalized rainfall coefficient, the rain cloud influence coefficient, and the vulnerability coefficient to obtain the flood control coefficient.

[0049] Further, the flood control regulation model includes:

[0050] Classify the flood control coefficient into low risk level, medium risk level, and high risk level;

[0051] Take minimizing the flood control coefficient and flood control cost as the fitness function, and use the genetic algorithm for the real-time regulation according to the fitness function and scheduling constraints; wherein, the real-time regulation includes regular regulation and emergency regulation;

[0052] If the flood control coefficient is at the low risk level, perform the regular regulation; if the flood control coefficient is at the medium risk level, perform the regular regulation and issue a warning; if the flood control coefficient is at the high risk level, adjust the weight of the fitness function and perform the emergency regulation.

[0053] Further, the scheduling constraints include:

[0054] Flood regulation maximum and minimum water level constraints, which are used to keep the reservoir water level within the allowable maximum and minimum ranges;

[0055] Discharge maximum and minimum flow constraints, which are used to keep the reservoir flow within the allowable maximum and minimum ranges;

[0056] Outflow change constraint of the reservoir, which is used to keep the change range of the reservoir outflow within the allowable range;

[0057] Water balance constraint, which is used to update the reservoir water storage according to the reservoir outflow and inflow;

[0058] Multi-reservoir coordination constraint, which is used to coordinate the outflow and inflow of the upstream and downstream reservoirs.

[0059] Compared with the prior art, the beneficial effects of the present invention are:

[0060] 1. By using a CNN model for image recognition of rain cloud regions, the spatial features of rain clouds can be extracted, thereby calculating the edge roughness coefficient, internal density coefficient, and rain cloud thickness coefficient, improving the accuracy of rainfall prediction and making weather monitoring more reliable. In addition, by using the optical flow method to analyze the moving speed and direction of rain clouds, the dynamics of rain clouds can be grasped in real time, and its potential impact on the water conservancy project system can be evaluated in a timely manner, providing a scientific basis for scheduling decisions, thus effectively reducing the flood risk and improving the flexibility of real-time scheduling of the water conservancy project system.

[0061] 2. By using a deep learning model to predict the future water level of a reservoir, the flood risk can be accurately evaluated. Calculating the flood risk coefficient in combination with historical flood defense data helps to quantify the losses that floods may bring. By analyzing the topographic vulnerability coefficient related to terrain and the social vulnerability coefficient related to building and population density, the sensitivity of the system to floods can be comprehensively reflected. Multi-dimensional evaluation can help formulate more targeted flood control strategies, improve the ability to respond to floods, and thus improve the flexibility of real-time scheduling of the water conservancy project system.

[0062] 3. By calculating the flood control coefficient and establishing a flood control regulation model based on the flood control coefficient, different risk levels are effectively distinguished, ensuring that strategies can be quickly adjusted and scheduling can be executed when facing different risks, thereby enhancing the flexibility and response speed in dealing with sudden flood events. In addition, combined with scheduling constraints such as water level, flow rate, and water volume balance, the safety and effectiveness of the scheduling process are ensured. This real-time regulation method not only improves the scientificity and flexibility of water conservancy project system management, but also reduces the flood risk and potential losses. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a schematic flowchart of the method provided by the embodiment of the present invention;

[0064] Figure 2 It is a schematic flowchart of the CNN model training provided by the embodiment of the present invention;

[0065] Figure 3 It is a schematic structural diagram of the influence coefficient calculation provided by the embodiment of the present invention;

[0066] Figure 4 It is a schematic structural diagram of the flood control coefficient calculation provided by the embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0067] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0068] Please refer to Figures 1 to 4 , the present invention provides a real-time scheduling method for a water conservancy project system based on neural network technology, and the technical solution is as follows:

[0069] With the frequent occurrence of climate change and extreme weather events, the regulation of water conservancy project systems has become more complex. Especially in the rainy season, it often faces the flood risk of the superposition of local heavy rain and upstream inflow. Traditional scheduling methods relying on historical data and experience are difficult to cope with sudden hydrological changes, and cannot achieve rapid scheduling of real-time rain cloud dynamics, water levels, and river basins, resulting in low efficiency in flood control and water resource management. Although existing real-time scheduling methods have improved the response ability, they lack a comprehensive analysis of rain cloud dynamics, terrain, and social factors, resulting in limited flexibility of scheduling strategies.

[0070] To address these problems, a real-time scheduling method for a water conservancy project system based on neural network technology is proposed. This method can accurately predict rainfall, flood trends, and future water level changes by combining deep learning models with real-time acquisition of information such as satellite cloud images, rainfall data, reservoir water levels, and river basin flows. In addition, this method also formulates corresponding scheduling plans according to different flood control risk levels, realizes flexible scheduling under low, medium, and high risk levels, ensures the reasonable allocation of regional water use, improves water resource utilization efficiency, avoids unnecessary flood discharge or water resource waste, and thus enhances the flexibility of system regulation.

[0071] Embodiment 1

[0072] Region A is facing the water resource pressure brought by the increasing population and economic activities, and is also affected by the frequent occurrence of extreme rainfall events caused by climate change, resulting in an increased flood risk. The original water conservancy project system cannot respond in a timely manner to sudden floods and water resource allocation. For the water conservancy project system in Region A, a real-time scheduling method for a water conservancy project system based on neural network technology is proposed, aiming to improve the flexibility of system real-time regulation, as Figure 1 shown, including:

[0073] Obtain real-time data of the water conservancy project system at fixed time intervals. The real-time data includes satellite cloud images, vertical rain cloud profile data, and flood control data, and preprocess the satellite cloud images to obtain a first image dataset.

[0074] Furthermore, the flood control data includes: flood defense historical data, reservoir historical data, reservoir real-time data, terrain and geographical data, building density, population density, and hydrological facilities.

[0075] Specifically, the fixed time interval can be set to 30 minutes. Meteorological satellites are used to transmit satellite cloud images in real-time, and ground receiving stations or specialized meteorological service platforms are combined to obtain satellite cloud images. Equipment such as ground radars and meteorological sounding balloons are used to obtain vertical profile data of rain clouds, including cloud thickness and density at different heights, etc., so as to timely grasp the latest situation of rain clouds and enhance the dynamic response ability of flood control and water resource scheduling. Flood control data is obtained through historical databases, basin monitoring stations, and online sensing systems. The flood defense historical data includes past flood events, maximum flood peak flows, flood impact ranges, and loss situations, etc. The reservoir historical data includes changes in reservoir water storage, annual scheduling records, and flood peak response measures, etc. The reservoir real-time data includes real-time water levels, inflow and outflow discharges, current water storage, and flood discharge status, etc. The terrain and geographical data are obtained through remote sensing technology or previous engineering records. The building density, population density, and hydrological facilities are obtained through satellite images, urban planning data, and water conservancy department databases. Through the collection of these data, the impact of floods on infrastructure and residents within the region can be comprehensively evaluated, the current rain cloud, basin, and reservoir conditions can be comprehensively analyzed, providing an accurate basis for flood control scheduling, thereby enhancing the flexibility of system regulation.

[0076] Furthermore, the preprocessing includes:

[0077] Perform color space conversion on the satellite cloud image to obtain the first cloud image;

[0078] Specifically, the original format of the satellite cloud image adopts RGB or other color spaces. By converting the image into a color space suitable for cloud recognition and analysis, cloud features can be better highlighted. For example, converting an RGB satellite image containing clouds of different thicknesses into the HSV space can enhance the brightness difference of the clouds, facilitating subsequent recognition algorithm processing.

[0079] Correct the illumination and radiation of the first cloud image to obtain the second cloud image;

[0080] Specifically, due to the influence of illumination conditions and atmospheric radiation during satellite imaging, the clouds in the satellite image may be disturbed by uneven illumination. For example, in the satellite image, the clouds in the east are brighter, while the west is darker due to the sun's angle. Through correction, the brightness of the entire image can be made uniform, reducing the error caused by light differences.

[0081] Determine the downsampling ratio according to the task requirements, and perform downsampling on the second cloud map according to the downsampling ratio using bilinear interpolation to obtain the third cloud map, which can reduce the number of pixels while maintaining the clarity and details of the image as much as possible.

[0082] Perform noise removal on the third cloud map using Gaussian filtering to obtain the first image dataset, which can smooth the image and make the edges and structures of the clouds clearer during subsequent image detection.

[0083] By preprocessing the image, not only does the satellite cloud map become of higher quality, less noisy, and easier to process, but it also improves the system processing speed and analysis accuracy, thereby enhancing the reliability and flexibility of flood control coefficient calculation and water conservancy scheduling.

[0084] Furthermore, use a CNN model to perform image recognition on the first image dataset. The CNN model is used to extract the spatial features of rain clouds to obtain a rain cloud region map and a rain cloud region edge map. The image recognition process is as follows: perform feature extraction based on the first image dataset to obtain low-level image features; form a first branch network and a second branch network based on the low-level image features. The first branch network is used to segment the rain cloud region and output the rain cloud region map, and the second branch network is used to extract the edge of the rain cloud region and output the rain cloud region edge map;

[0085] Furthermore, as Figure 2 shown, the training steps of the CNN model include:

[0086] Step S1: Obtain historical satellite cloud maps and perform the preprocessing to obtain a historical cloud map dataset;

[0087] Step S2: Mark image labels for the historical cloud map dataset to obtain a second image dataset, and input the second image dataset into the CNN model in the form of [H, W, C], where H is the height of the image, W is the width, and C is the number of channels; among them, the image labels include rain cloud regions, non-rain cloud regions, edge points, and non-edge points;

[0088] Step S3: Perform a convolution operation on the second image dataset using a convolutional layer, slide a convolutional kernel with a size of 3×3 on the cloud map, and output a first feature map;

[0089] Step S4: Use the ReLU activation function to set the negative part of the convolution output to 0, and perform a pooling operation on the first feature map using a pooling layer to perform dimensionality reduction processing and then output a second feature map;

[0090] Step S5: Use the first branch network to perform a segmentation operation on the second feature map. Each pixel will be classified as "belonging to the rain cloud area" or "not belonging to the rain cloud area", and the rain cloud area map will be output;

[0091] Step S6: Use the second branch network to perform edge feature detection on the second feature map. Each pixel point is marked as an "edge point" or a "non-edge point", and the rain cloud area edge map is output;

[0092] Step S7: Use the cross-entropy loss function to calculate the segmentation loss of the first branch network, use the IoU Loss function to calculate the edge detection loss of the second branch network, and perform a weighted sum of the segmentation loss and the edge detection loss to obtain the joint loss value, expressed as:

[0093] ;

[0094] where, is the segmentation loss value, is the cross-entropy loss function, is the actual segmentation label, is the predicted output after model segmentation; is the edge loss value, is the IoU Loss function, is the actual label of the edge, is the predicted edge output; is the final loss value, and are hyperparameters of the weights of the two tasks and can be set to 0.5 and 0.5 respectively.

[0095] Step S8: According to the joint loss value, use the backpropagation algorithm to update the weights of the CNN model, and gradually improve the accuracy of rain cloud area recognition and edge detection;

[0096] Step S9: If the number of iterations reaches the preset number, stop training; otherwise, repeat Steps S3 to S9.

[0097] where, the preset number can be set to 500 times.

[0098] Through the multi-branch network structure of the CNN model, after 500 times of training, the final loss value is 0.2750, which can accurately identify the edge and area of the rain cloud, not only improving the processing efficiency of meteorological data, but also providing strong data support for the real-time scheduling and optimization of the water conservancy project system, thus enhancing the ability to respond to climate change and extreme weather events and the flexibility of system regulation.

[0099] Such as Figure 3As shown, the edge roughness coefficient and the internal density coefficient are calculated based on the rain cloud area map and the rain cloud area edge map; the rain cloud thickness coefficient is calculated based on the rain cloud vertical profile data, and the rainfall coefficient is calculated based on the edge roughness coefficient, the internal density coefficient, and the rain cloud thickness coefficient.

[0100] Further, the calculation process of the rainfall coefficient includes:

[0101] Using the box-counting method to calculate the fractal dimension of the edge of the rain cloud area edge map to obtain the edge roughness coefficient;

[0102] Specifically, convert the rain cloud area edge map into a binary image, with the edge point value being 1 and the rest being 0. Use multiple square boxes of different sizes, such as 1×1, 2×2 pixels, etc., and use grids of different sizes to cover the entire image. Calculate the number of boxes covered by at least one edge point, and use the fractal dimension formula to calculate the fractal dimension of the edge, expressed as:

[0103] ;

[0104] Where, is the fractal dimension, is the side length of the box, is the number of covered boxes; the fractal dimension is used to describe the complexity of the shape. The larger the fractal dimension, the more complex the rain cloud edge and the higher the roughness.

[0105] The edge roughness coefficient is calculated from the fractal dimension of the edge, expressed as:

[0106] ;

[0107] Where, is the edge roughness coefficient, is the minimum fractal dimension (such as a straight line) among all rain cloud edges, is the maximum fractal dimension (such as a fractal graph) among all rain cloud edges.

[0108] Calculate the variance of the gray values of the pixel points of the rain cloud area map to obtain the internal density coefficient.

[0109] Specifically, if the rain cloud area map is a color image, it needs to be converted into a gray image and the variance of the gray values is calculated, expressed as:

[0110] ;

[0111] Where, is the total number of pixels in the rain cloud area, is the gray value of the i-th pixel, is the mean value of the gray values, is the variance of the gray value;

[0112] Obtain the internal density coefficient based on the variance of the gray value, expressed as:

[0113] ;

[0114] wherein, is the internal density coefficient, is the maximum variance of the gray value;

[0115] Calculate the vertical height difference between the cloud top and the cloud bottom using the rain cloud vertical profile data, calculate the thickness change rate for the vertical height differences of consecutive time intervals, and obtain the rain cloud thickness coefficient.

[0116] Specifically, obtain the cloud top height and the cloud bottom height using the rain cloud vertical profile data, and calculate the vertical height difference, expressed as:

[0117] ;

[0118] wherein, is the vertical height difference, is the cloud top height, is the cloud bottom height;

[0119] The thickness change rate is expressed as:

[0120] ;

[0121] wherein, is the thickness change rate, is the vertical height difference at time is the vertical height difference at time;

[0122] After normalizing the thickness change rate, obtain the rain cloud thickness coefficient, expressed as:

[0123] ;

[0124] wherein, is the rain cloud thickness coefficient, is the maximum change rate of the height difference;

[0125] Perform a weighted sum of the edge roughness coefficient, the internal density coefficient, and the rain cloud thickness coefficient to obtain the rainfall coefficient, expressed as:

[0126] ;

[0127] wherein, is the rainfall coefficient, , and are the weight parameters of each coefficient, which can be set to 0.3, 0.3, and 0.4. The higher the complexity of the edge, the stronger the meteorological activity. By calculating the edge roughness coefficient, the cloud edge morphology information can be provided as a potential indicator of rainfall intensity and duration. The internal density coefficient reflects the concentration and distribution characteristics of water vapor in the cloud, further helping to analyze the rainfall possibility. In addition, the rate of change of the rain cloud thickness can evaluate the vertical thickness change of the rain cloud to capture its morphological evolution. Through the comprehensive evaluation of these three coefficients, the rainfall possibility can be described more accurately, thereby improving the flexibility of system regulation.

[0128] The optical flow method is used to calculate the moving speed and moving direction of the rain cloud for the rain cloud area maps at continuous time intervals, and the shortest distance between the central rain cloud and the water conservancy project area within the rain cloud area map is calculated; the rain cloud influence coefficient is calculated based on the rain cloud moving speed, the rain cloud moving direction, and the shortest distance.

[0129] Furthermore, the calculation process of the rain cloud influence coefficient includes:

[0130] The optical flow method is used to calculate the motion vector field of each pixel point for the rain cloud area maps at continuous time intervals. The direction of the motion vector field is the rain cloud moving direction, and the average modulus length of the motion vector field is the rain cloud moving speed;

[0131] According to the rain cloud coordinates and the water conservancy project area coordinates, the shortest distance is calculated using the Euclidean formula, expressed as:

[0132] ;

[0133] where is the shortest distance, is the central coordinate of the rain cloud, is the central coordinate of the water conservancy project area;

[0134] The cosine value of the angle between the rain cloud moving direction and the direction from the central rain cloud to the water conservancy project area is calculated, and the cosine value is multiplied by the rain cloud moving speed to obtain the second rain cloud moving speed; the shortest distance is divided by the second rain cloud moving speed to calculate the rain cloud influence coefficient, expressed as:

[0135] ;

[0136] where is the rain cloud influence coefficient, is the rain cloud moving speed, is the cosine value of the angle between the rain cloud moving direction and the direction of the water conservancy project area, It is the angle between the moving direction of the rain cloud and the direction of the water conservancy project area. By calculating the rain cloud influence coefficient, the potential influence of the rain cloud on the water conservancy project area can be dynamically reflected. Especially when dealing with sudden rainfall, it can improve the flexibility of system regulation, thus enhancing the early warning ability and dispatching response efficiency of the system.

[0137] Using the flood control data, the future water level of the reservoir is predicted by a deep learning model, and based on the flood control data and the future water level of the reservoir, a vulnerability coefficient is calculated.

[0138] Furthermore, the calculation process of the vulnerability coefficient includes:

[0139] Input the historical data of the reservoir into the basic model for training to obtain a trained deep learning model; wherein, the historical data of the reservoir includes historical water level data, meteorological data, and the inflow and outflow of the reservoir.

[0140] Input the real-time data of the reservoir into the deep learning model for prediction to obtain the future water level of the reservoir.

[0141] Among them, the deep learning model can be an LSTM model, a GRU model, or a CNN model.

[0142] Specifically, in the training stage, the real water level is used as the label, and the mean square error is used as the loss function to measure the deviation between the actual water level and the predicted water level. The model continuously adjusts the parameters until the error between its prediction result and the actual water level data reaches the lowest. After 1000 times of training, the error remains at 0.17. In the prediction stage, the model can output the water level change in the future for a period of time through calculation. For example, the water level will reach 150.5 meters tomorrow.

[0143] Calculate the ratio of the future water level of the reservoir to the reservoir storage capacity to obtain the reservoir storage coefficient.

[0144] Specifically, for example, the maximum storage capacity of the reservoir is 50 million cubic meters, and the water storage volume corresponding to the predicted water level is 30 million cubic meters, then the reservoir storage coefficient is 0.6.

[0145] Obtain the number of flood occurrences and flood losses from the flood defense historical data to get the flood occurrence frequency and flood loss rate, and multiply the flood occurrence frequency by the flood loss rate to obtain the flood risk coefficient.

[0146] Specifically, for example, in a certain area's water conservancy project area in history, 5 floods occurred within 20 years, with a total loss of 50 million yuan. The total economic value in 20 years was 20 billion yuan. After calculation, the flood risk coefficient can be obtained as 0.000625. If the calculated value is too small, the value can be amplified according to the theoretical upper limit of the risk coefficient of this area, that is, the product of the maximum total flood loss and the maximum flood loss rate. For example, if the theoretical upper limit is 0.01, then the flood risk coefficient is 0.0625.

[0147] Obtain the slope and soil type based on topographic and geographical data, and divide grades according to the slope and the soil type to obtain a topographic score; calculate the average topographic score of the water conservancy project area based on the topographic score to obtain a topographic vulnerability coefficient;

[0148] Specifically, the division method can be:

[0149] If the slope is less than 5%, or the soil type is of low vulnerability, a score of 0.3 is assigned; if the slope is less than 15% and greater than 5%, or the soil type is of medium vulnerability, a score of 0.6 is assigned; if the slope is greater than 15%, or the soil type is of high vulnerability, a score of 1 is assigned.

[0150] Perform a weighted sum of the building density and the population density to obtain a social vulnerability coefficient;

[0151] Among them, the building density and the population density are respectively normalized, and the weights can be set to 0.4 and 0.6 respectively.

[0152] Perform a weighted sum of the reservoir water storage coefficient, the flood risk coefficient, the topographic vulnerability coefficient, and the social vulnerability coefficient to obtain the vulnerability coefficient, expressed as:

[0153] ;

[0154] Among them, is the vulnerability coefficient, is the reservoir water storage coefficient, is the flood risk coefficient, is the topographic vulnerability coefficient, is the social vulnerability coefficient, 、 、 and are the weights of each coefficient respectively, and can all be set to 0.2.

[0155] Predicting the future water level of a reservoir through a deep learning model, calculating the vulnerability coefficient by combining flood control data, terrain data, and social data, can provide potential risk information of the reservoir, help the water conservancy department take measures in advance, improve flood control and disaster resistance capabilities, ensure the safety of the reservoir and its surrounding areas, and thus enhance the scientific nature of reservoir management and the accuracy of real-time scheduling.

[0156] Calculate the flood control coefficient according to the rainfall coefficient, the rain cloud influence coefficient, and the vulnerability coefficient, and establish a flood control regulation model according to the flood control coefficient. The flood control regulation model is used for real-time regulation of the water conservancy project system.

[0157] Furthermore, as Figure 4 shown, the calculation of the flood control coefficient includes:

[0158] Normalize the rainfall coefficient, the rain cloud influence coefficient, and the vulnerability coefficient respectively, and perform weighted summation on the normalized rainfall coefficient, the rain cloud influence coefficient, and the vulnerability coefficient to obtain the flood control coefficient, expressed as:

[0159] ;

[0160] Among them, is the flood control coefficient, is the rainfall coefficient, is the rain cloud influence coefficient, is the vulnerability coefficient, 、 and are the weights of each coefficient respectively, and can be set to 0.4, 0.3, and 0.3.

[0161] By combining rainfall intensity, rain cloud influence, and the vulnerability of the area, the calculation of the flood control coefficient is made more accurate, and the possible flood risk can be predicted more accurately, thus improving the flexibility of system regulation. In addition, by setting weights, it can be adjusted according to different regional characteristics, climate patterns, and infrastructure. For example, in areas with more rainfall, the weight of the rainfall coefficient can be increased, while in areas with higher social vulnerability, more attention can be paid to the vulnerability coefficient, so as to better adapt to flood control plans in different regions.

[0162] Furthermore, the flood control regulation model includes:

[0163] Classify the flood control coefficient into low-risk level, medium-risk level, and high-risk level;

[0164] Specifically, the classification of risk levels can be set according to the value range of the flood control coefficient. Here, the value range of the flood control coefficient is 0-1, and the risk levels are:

[0165] Low-risk level ( ): The reservoir is operating normally with a low risk.

[0166] Medium risk level ( ): The reservoir has a certain flood risk and requires an appropriate increase in drainage volume.

[0167] High risk level ( ): The reservoir faces a high flood risk and immediate emergency measures must be taken, such as quickly discharging water and increasing the scheduling frequency.

[0168] Minimize the flood control coefficient and flood control cost as the fitness function, and use a genetic algorithm to perform the real-time regulation according to the fitness function and scheduling constraints; wherein, the real-time regulation includes regular regulation and emergency regulation;

[0169] If the flood control coefficient is at the low risk level, perform the regular regulation; if the flood control coefficient is at the medium risk level, perform the regular regulation and issue a warning; if the flood control coefficient is at the high risk level, adjust the weight of the fitness function and perform the emergency regulation.

[0170] Specifically, the fitness function can be expressed as:

[0171] ;

[0172] Wherein, is the fitness, is the normalized flood control cost, and are the weight coefficients of the two objectives respectively.

[0173] Among them, at the high risk level, increase the priority of emergency regulation and reduce the weight of cost, and can be set to 0.95, can be set to 0.05; at the low risk level and medium risk level, can be set to 0.7, can be set to 0.3.

[0174] To ensure that the scheduling scheme meets all constraint conditions, the violations of constraints can be added to the fitness function, and the extended fitness function is:

[0175] ;

[0176] Wherein, is a penalty function corresponding to each constraint condition, and the greater the degree of violation, the greater the penalty.

[0177] Specifically, the implementation steps of the genetic algorithm are:

[0178] First, an initial population is generated with a population size of 30, and each chromosome represents a reservoir operation plan. The chromosome includes operation parameters such as the discharge flow and inflow at each time step;

[0179] Calculate the fitness function of each chromosome. The higher the fitness value, the more compliant the operation plan represented by the chromosome is;

[0180] Use roulette wheel selection to select chromosomes to generate a new generation of chromosomes. The higher the fitness, the greater the chance of a chromosome being selected;

[0181] Exchange the operation parameters of two chromosomes at different time periods to generate new chromosomes, and the probability can be set to 0.8;

[0182] Randomly adjust the operation parameters in some time periods to increase the diversity of the population, and the probability can be set to 0.05;

[0183] Generate a new population through operations such as selection, crossover, and mutation;

[0184] If the number of iterations reaches 800 times, terminate the iteration and output the operation parameters.

[0185] By taking the minimization of the flood control coefficient and flood control cost as the fitness function, both flood control safety and economic cost are taken into account. In addition, by combining the conventional regulation and emergency regulation plans, real-time adjustment can be made according to the flood control coefficient, improving the flexibility and safety of operation and avoiding the risks of extreme weather or sudden floods. Finally, the effective combination of hierarchical scheduling and genetic algorithm realizes the efficient and safe regulation of the water conservancy system, improving the flexibility of system regulation.

[0186] Furthermore, the operation constraints include:

[0187] Flood control maximum and minimum water level constraints, used to keep the reservoir water level within the allowed maximum and minimum ranges, expressed as:

[0188] ;

[0189] where is the water level at time t, is the minimum allowed water level, which can be set to 40m, is the maximum allowed water level, which can be set to 80m.

[0190] Discharge maximum and minimum flow constraints, used to keep the reservoir flow within the allowed maximum and minimum ranges, expressed as:

[0191] ;

[0192] where is the discharge flow at time t, is the minimum discharge flow rate, which can be set to 100 cubic meters per second, is the maximum discharge flow rate, which can be set to 400 cubic meters per second.

[0193] The discharge flow rate change constraint is used to keep the change range of the reservoir discharge flow rate within the allowable range, expressed as:

[0194] ;

[0195] where, is the flow rate change between adjacent time steps, is the discharge flow rate at time t - 1, is the maximum allowable flow rate change, which can be set to 50 cubic meters per second.

[0196] The water balance constraint is used to update the reservoir water storage according to the reservoir discharge flow rate and the inflow rate, expressed as:

[0197] ;

[0198] where, is the reservoir storage volume at time t + 1, is the reservoir storage volume at time t, is the inflow rate at time t.

[0199] The multi - reservoir coordination constraint is used to coordinate the discharge flow rate and the inflow rate of the upstream and downstream reservoirs, expressed as:

[0200] ;

[0201] where, is the discharge flow rate of the upstream reservoir at time t, is the inflow rate of the current reservoir at time t, is the inflow carrying capacity of the downstream reservoir, which can be set to 300 cubic meters per second.

[0202] Introducing these constraint conditions in real - time scheduling not only strengthens the cooperation between upstream and downstream reservoirs, improves the regional flood control ability, but also ensures that the water level and flow rate of the reservoir are kept within a safe range, effectively suppresses the drastic fluctuation of the discharge flow rate, and avoids the risks of overflow and drying up, thus improving the flexibility of regulation.

[0203] Example 2

[0204] Region B has abundant water resources. However, due to its complex terrain and diverse climate, the management of water resources faces challenges. With the impact of climate change, Region B has also experienced extreme weather events such as droughts and heavy rains, resulting in uneven water resource distribution and ecosystem degradation. Therefore, the water conservancy project system in Region B requires a new regulation method to improve the flexibility of regulation. A real-time scheduling method for a water conservancy project system based on neural network technology includes:

[0205] Obtain real-time data of the water conservancy project system at fixed time intervals. The real-time data includes satellite cloud images, vertical cross-section data of rain clouds, and flood control data. Preprocess the satellite cloud images to obtain a first image dataset;

[0206] Among them, the time interval can be set to 1h;

[0207] Use a CNN model to perform image recognition on the first image dataset. The CNN model is used to extract the spatial features of rain clouds to obtain a rain cloud area map and a rain cloud area edge map. The image recognition process is as follows: perform feature extraction according to the first image dataset to obtain low-level image features; form a first branch network and a second branch network according to the low-level image features. The first branch network is used to segment the rain cloud area and output the rain cloud area map, and the second branch network is used to extract the edge of the rain cloud area and output the rain cloud area edge map;

[0208] Calculate the edge roughness coefficient and the internal density coefficient according to the rain cloud area map and the rain cloud area edge map; calculate the rain cloud thickness coefficient according to the vertical cross-section data of the rain cloud, and calculate the rainfall coefficient according to the edge roughness coefficient, the internal density coefficient, and the rain cloud thickness coefficient;

[0209] Use the optical flow method to calculate the moving speed and moving direction of the rain cloud for consecutive time intervals of the rain cloud area map, and calculate the shortest distance between the central rain cloud in the rain cloud area map and the water conservancy project area; calculate the rain cloud influence coefficient according to the rain cloud moving speed, the rain cloud moving direction, and the shortest distance;

[0210] Use a deep learning model to predict the future water level of the reservoir according to the flood control data, and calculate according to the flood control data and the future water level of the reservoir to obtain the vulnerability coefficient;

[0211] Calculate the flood control coefficient according to the rainfall coefficient, the rain cloud influence coefficient, and the vulnerability coefficient, and establish a flood control regulation model according to the flood control coefficient. The flood control regulation model is used to perform real-time regulation on the water conservancy project system.

[0212] Furthermore, the flood control data includes: flood defense historical data, reservoir historical data, reservoir real-time data, terrain and geographical data, building density, population density, and hydrological facilities.

[0213] Furthermore, the preprocessing includes:

[0214] Obtain the historical satellite cloud map data of area B in the past five years, perform color space conversion on the satellite cloud map, convert the original cloud map from the RGB color space to the HSV color space to obtain the first cloud map;

[0215] Correct the illumination and radiation of the first cloud map to ensure the illumination consistency of different regions in the image to obtain the second cloud map;

[0216] Determine the downsampling ratio according to the task requirements. Here, the downsampling ratio is set to 0.5, and bilinear interpolation is used to downsample the second cloud map according to the downsampling ratio to reduce the computational burden and obtain the third cloud map;

[0217] Use Gaussian filtering to remove noise from the third cloud map, set the filter size to 5×5 to improve the image quality and obtain the first image dataset.

[0218] Furthermore, the training steps of the CNN model include:

[0219] Step S1: Obtain historical satellite cloud maps and perform the preprocessing to obtain a historical cloud map dataset;

[0220] Step S2: Mark image labels for the historical cloud map dataset to obtain a second image dataset, and input the second image dataset into the CNN model; among them, the image labels include rain cloud regions, non-rain cloud regions, edge points, and non-edge points for subsequent model training;

[0221] Step S3: Perform a convolution operation on the second image dataset using a convolutional layer, set the convolutional kernel size to 3×3, the stride to 1, and output the first feature map;

[0222] Step S4: Perform a pooling operation on the first feature map using the ReLU activation function and a pooling layer, with a pooling window of 2×2 and a stride of 2, and output the second feature map;

[0223] Step S5: Use the first branch network to perform a segmentation operation on the second feature map and output the rain cloud region map;

[0224] Step S6: Use the second branch network to perform edge feature detection on the second feature map and output the rain cloud region edge map;

[0225] Among them, the first branch network and the second branch network are fully convolutional networks;

[0226] Step S7: Use the cross-entropy loss function to calculate the segmentation loss of the first branch network, use the IoU Loss function to calculate the edge detection loss of the second branch network, and perform weighted summation on the segmentation loss and the edge detection loss to obtain a joint loss value;

[0227] Among them, the segmentation loss weight is set to 0.7, and the edge detection loss weight is set to 0.3.

[0228] Step S8: Update the CNN model according to the joint loss value;

[0229] Step S9: If the number of iterations reaches the preset number of times, and the maximum number of training iterations is set to 1000 times, then stop training; otherwise, repeat steps S3 to S9.

[0230] Furthermore, the calculation process of the rainfall coefficient includes:

[0231] Use the box-counting method to calculate the fractal dimension of the edge of the rain cloud area edge map to obtain the edge roughness coefficient;

[0232] Among them, the edge roughness coefficient can be directly equal to the fractal dimension of the edge;

[0233] Calculate the variance of the pixel grayscale values of the rain cloud area map to obtain the internal density coefficient;

[0234] Specifically, assume that the grayscale values of the rain cloud area map are [120, 130, 140, 135, 125, 128, 132, 138, 142, 134], calculate the average grayscale value to be 131.5, then the grayscale value variance is 50.25.

[0235] Use the rain cloud vertical profile data to calculate the vertical height difference between the cloud top and the cloud bottom, and calculate the thickness change rate for the vertical height differences at consecutive time intervals to obtain the rain cloud thickness coefficient;

[0236] Specifically, assume that the cloud top height is 3000m and the cloud bottom height is 1500m, then the vertical height difference is 1500m. If the height differences before and after within 1 hour are 1500m and 1200m respectively, then the change rate is 300m / h.

[0237] Normalize the edge roughness coefficient, the internal density coefficient, and the rain cloud thickness coefficient and then perform weighted summation to obtain the rainfall coefficient.

[0238] Among them, the weights are set to 0.4, 0.4, and 0.2 respectively.

[0239] Furthermore, the calculation process of the rain cloud influence coefficient includes:

[0240] Using the optical flow method to calculate the motion vector field of each pixel point for the rain cloud area maps of consecutive time intervals. The direction of the motion vector field is the moving direction of the rain cloud, and the average modulus length of the motion vector field is the moving speed of the rain cloud.

[0241] Specifically, the motion vector of each pixel point is expressed as (dx, dy), which are the change amounts in the horizontal and vertical directions respectively. Then the calculation formula for the average modulus length is:

[0242] ;

[0243] Where is the average modulus length, is the number of pixel points, and i is the pixel point serial number;

[0244] According to the rain cloud coordinates and the coordinates of the water conservancy project area, use the Euclidean formula to calculate the obtained shortest distance;

[0245] Calculate the cosine value of the angle between the moving direction of the rain cloud and the direction from the central rain cloud to the water conservancy project area, and multiply the cosine value by the moving speed of the rain cloud to obtain the second rain cloud moving speed;

[0246] Divide the shortest distance by the second rain cloud moving speed to calculate the obtained rain cloud influence coefficient.

[0247] Furthermore, the calculation process of the vulnerability coefficient includes:

[0248] Input the historical data of the reservoir in the past 5 years into the basic model for training to obtain a trained deep learning model, and the prediction accuracy rate of the model on the test set reaches 87%; among them, the historical data of the reservoir includes historical water level data, meteorological data, and reservoir inflow and outflow;

[0249] Input the real-time data of the reservoir into the trained deep learning model for prediction to obtain the future water level of the reservoir;

[0250] Calculate the ratio of the future water level of the reservoir to the reservoir storage capacity to obtain the reservoir storage coefficient;

[0251] Obtain the number of flood occurrences and flood losses according to the historical data of flood prevention to get the flood occurrence frequency and flood loss rate, and multiply the flood occurrence frequency by the flood loss rate to obtain the flood risk coefficient;

[0252] Obtain the slope and soil type based on the terrain and geographical data, divide the grades according to the slope and the soil type to obtain the terrain score; calculate the average terrain score of the water conservancy project area according to the terrain score to obtain the terrain vulnerability coefficient;

[0253] Perform weighted summation of the building density and the population density to obtain the social vulnerability coefficient;

[0254] Perform weighted summation of the reservoir water storage coefficient, the flood risk coefficient, the terrain vulnerability coefficient and the social vulnerability coefficient to obtain the vulnerability coefficient.

[0255] Among them, the weights are set to 0.25, 0.35, 0.2 and 0.2.

[0256] Furthermore, the calculation of the flood control coefficient includes:

[0257] Normalize the rainfall coefficient, the rain cloud influence coefficient and the vulnerability coefficient respectively, and perform weighted summation of the normalized rainfall coefficient, the rain cloud influence coefficient and the vulnerability coefficient to obtain the flood control coefficient, which provides a quantitative basis for flood control management in Area B and is convenient for formulating effective flood control measures and emergency plans.

[0258] Among them, the weights are set to 0.4, 0.3 and 0.3.

[0259] Furthermore, the flood control regulation model includes:

[0260] Divide the flood control coefficient into grades, including low risk grade, medium risk grade and high risk grade;

[0261] Among them, the risk grades are: low risk grade ( ), medium risk grade ( ), high risk grade ( ). Assume that the calculated flood control coefficient is 0.6982, then it is within the medium risk grade range.

[0262] Take the minimization of the flood control coefficient and the flood control cost as the fitness function, and perform the real-time regulation using the genetic algorithm according to the fitness function and the scheduling constraints;

[0263] Among them, the water conservancy real-time regulation includes conventional regulation and emergency regulation; the parameter settings of the genetic algorithm are: the population size is 100, the selection probability is 40%, the crossover probability is 80%, the mutation probability is 5%, and the termination condition is the maximum number of iterations of 1000 generations.

[0264] If the flood control coefficient is at the low - risk level, the conventional regulation is executed; if the flood control coefficient is at the medium - risk level, the conventional regulation is executed and a warning is issued; if the flood control coefficient is at the high - risk level, the weight of the fitness function is adjusted and the emergency regulation is executed.

[0265] Among them, when at the high - risk level, the flood control coefficient is set to 0.9 and the cost weight is set to 0.1; when at the low - risk level and the medium - risk level, the flood control coefficient is set to 0.5 and the cost weight is set to 0.5.

[0266] Furthermore, the scheduling constraints include:

[0267] The maximum and minimum flood - regulation water - level constraints, which are used to keep the reservoir water - level within the permitted maximum and minimum ranges;

[0268] Among them, the maximum water - level of the reservoir is set to 100 meters and the minimum water - level is set to 50 meters.

[0269] The maximum and minimum discharge - flow constraints, which are used to keep the reservoir discharge - flow within the permitted maximum and minimum ranges;

[0270] Among them, the maximum discharge - flow is 500 cubic meters per second and the minimum discharge - flow is 100 cubic meters per second.

[0271] The constraint on the change of the outflow - flow of the reservoir, which is used to keep the change range of the outflow - flow of the reservoir within the permitted range;

[0272] Among them, the change range of the outflow - flow should be kept within 50 cubic meters per second.

[0273] The water - balance constraint, which is used to update the water storage of the reservoir according to the outflow - flow and inflow - flow of the reservoir;

[0274] Among them, the water storage is updated according to the inflow - flow of the reservoir (such as 600 cubic meters per second) and the outflow - flow of the reservoir (such as the current outflow - flow is 300 cubic meters per second).

[0275] The multi - reservoir coordination constraint, which is used to coordinate the outflow - flow and inflow - flow of the upstream and downstream reservoirs.

[0276] Among them, the inflow - carrying capacity of the downstream reservoir is set to 250 cubic meters per second.

[0277] Specifically, assuming that the flood control coefficient is at the medium - risk level, the system executes the conventional regulation, issues a warning, notifies relevant personnel to pay attention to the possible flood risk, and then calculates the scheduling parameters using the genetic algorithm. For example, the discharge - flow is adjusted to 400 cubic meters per second. By combining the genetic algorithm to optimize the real - time scheduling strategy of the water conservancy system, in the case of the intensification of flood risk, it can provide effective decision - making support for water conservancy management using the scheduling constraints, and improve the flexibility of system regulation.

[0278] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A real-time dispatching method for a water conservancy project system based on neural network technology, comprising: Acquire real-time data of a water conservancy project system at fixed time intervals, the real-time data including satellite cloud images, rain cloud vertical profile data and flood control data, and pre-process the satellite cloud images to obtain a first image data set; Performing image recognition on the first image data set using a CNN model, wherein the CNN model is used to extract spatial features of rain clouds to obtain a rain cloud area map and a rain cloud area edge map; The training steps of the CNN model include: Step S1: Acquire historical satellite cloud images and perform the preprocessing to obtain a historical cloud image data set; Step S2: labeling the image labels of the historical cloud image dataset to obtain a second image dataset, and inputting the second image dataset into the CNN model; wherein the CNN model includes a first branch network and a second branch network; the image labels include rain cloud areas, non-rain cloud areas, edge points, and non-edge points; Step S3: using a convolution layer to perform a convolution operation on the second image data set, and output a first feature map; Step S4: using a ReLU activation function and a pooling layer to perform a pooling operation on the first feature map, and output a second feature map; Step S5: using the first branch network to perform a segmentation operation on the second feature map, and outputting the rain cloud area map; Step S6: using the second branch network to perform edge feature detection on the second feature map, and outputting the rain cloud area edge map; Step S7: using the cross entropy loss function to calculate the segmentation loss of the first branch network, using the IoU Loss loss function to calculate the edge detection loss of the second branch network, performing weighted summation of the segmentation loss and the edge detection loss to obtain a joint loss value; Step S8: updating the CNN model according to the joint loss value; Step S9: If the number of iterations reaches the preset number, stop training, otherwise repeat steps S3 to S9; An edge roughness coefficient and an internal density coefficient are calculated based on the rain cloud area map and the rain cloud area edge map; a rain cloud thickness coefficient is calculated based on the rain cloud vertical profile data; a rainfall coefficient is calculated based on the edge roughness coefficient, the internal density coefficient and the rain cloud thickness coefficient; The rain cloud area map at continuous time intervals is calculated using the optical flow method to obtain the rain cloud moving speed and the rain cloud moving direction, and the shortest distance between the central rain cloud in the rain cloud area map and the water conservancy project area is calculated; the rain cloud influence coefficient is calculated according to the rain cloud moving speed, the rain cloud moving direction and the shortest distance; The calculation process of the rain cloud influence coefficient includes: The optical flow method is used to calculate the motion vector field of each pixel point for the rain cloud area map at continuous time intervals, wherein the direction of the motion vector field is the moving direction of the rain cloud, and the average modulus length of the motion vector field is the moving speed of the rain cloud; The shortest distance is calculated using the Euclidean formula according to the coordinates of the rain cloud and the coordinates of the water conservancy project area; Calculating the cosine value of the angle between the rain cloud moving direction and the direction from the central rain cloud to the water conservancy project area, and multiplying the cosine value by the rain cloud moving speed to obtain a second rain cloud moving speed; Dividing the shortest distance by the second rain cloud moving speed to calculate the rain cloud influence coefficient; Predicting the future water level of the reservoir using a deep learning model based on the flood control data, and calculating a vulnerability coefficient based on the flood control data and the future water level of the reservoir; A flood control coefficient is calculated according to the rainfall coefficient, the rain cloud influence coefficient and the vulnerability coefficient, and a flood control model is established according to the flood control coefficient, wherein the flood control model is used to perform real-time control on the water conservancy project system; The flood control model includes: Classifying the flood control coefficient into levels, including low risk level, medium risk level and high risk level; Minimizing the flood control coefficient and the flood control cost is used as a fitness function, and a genetic algorithm is used to perform the real-time regulation according to the fitness function and the scheduling constraints; wherein the real-time regulation includes conventional regulation and emergency regulation; If the flood control coefficient is the low risk level, the conventional regulation is performed; if the flood control coefficient is the medium risk level, the conventional regulation is performed and a warning is issued; if the flood control coefficient is the high risk level, the weight of the fitness function is adjusted and the emergency regulation is performed.

2. A real-time dispatching method for water conservancy engineering system based on neural network technology according to claim 1, characterized in that: The flood control data include: historical flood defense data, historical reservoir data, real-time reservoir data, topographic and geographical data, building density, population density and hydrological facilities.

3. The real-time dispatching method of a water conservancy project system based on neural network technology according to claim 1 is characterized in that: The pre-processing comprises: Performing color space conversion on the satellite cloud image to obtain a first cloud image; Correcting the illumination and radiation of the first cloud image to obtain a second cloud image; Determine a downsampling ratio according to task requirements, and use bilinear interpolation to downsample the second cloud image according to the downsampling ratio to obtain a third cloud image; The third cloud image is subjected to a Gaussian filter to remove noise, thereby obtaining the first image data set.

4. The real-time dispatching method of a water conservancy project system based on neural network technology according to claim 1 is characterized in that: The training steps of the CNN model include: Step S1: Acquire historical satellite cloud images and perform the preprocessing to obtain a historical cloud image data set; Step S2: labeling the image labels of the historical cloud image dataset to obtain a second image dataset, and inputting the second image dataset into the CNN model; wherein the CNN model includes a first branch network and a second branch network; the image labels include rain cloud areas, non-rain cloud areas, edge points, and non-edge points; Step S3: using a convolution layer to perform a convolution operation on the second image data set, and output a first feature map; Step S4: using a ReLU activation function and a pooling layer to perform a pooling operation on the first feature map, and output a second feature map; Step S5: using the first branch network to perform a segmentation operation on the second feature map, and outputting the rain cloud area map; Step S6: using the second branch network to perform edge feature detection on the second feature map, and outputting the rain cloud area edge map; Step S7: using the cross entropy loss function to calculate the segmentation loss of the first branch network, using the IoU Loss loss function to calculate the edge detection loss of the second branch network, performing weighted summation of the segmentation loss and the edge detection loss to obtain a joint loss value; Step S8: updating the CNN model according to the joint loss value; Step S9: If the number of iterations reaches the preset number, stop training, otherwise repeat steps S3 to S9.

5. The real-time dispatching method of a water conservancy project system based on neural network technology according to claim 1 is characterized in that: The calculation process of the rainfall coefficient includes: Using a box counting method to calculate the fractal dimension of the edge of the rain cloud area edge map, to obtain the edge roughness coefficient; Calculating the variance of the grayscale values ​​of the pixels in the rain cloud area map to obtain the internal density coefficient; Calculating the vertical height difference between the cloud top and the cloud bottom using the rain cloud vertical profile data, calculating the thickness change rate for the vertical height difference at consecutive time intervals, and obtaining the rain cloud thickness coefficient; The edge roughness coefficient, the internal density coefficient and the rain cloud thickness coefficient are weightedly summed to obtain the rainfall coefficient.

6. The real-time dispatching method of a water conservancy project system based on neural network technology according to claim 1 is characterized in that: The calculation process of the vulnerability coefficient includes: Inputting the reservoir historical data into the basic model for training to obtain a trained deep learning model; wherein the reservoir historical data includes historical water level data, meteorological data, and reservoir inflow and outflow; Input the real-time data of the reservoir into the trained deep learning model for prediction to obtain the future water level of the reservoir; Calculate the ratio of the future water level of the reservoir to the water storage capacity of the reservoir to obtain the water storage coefficient of the reservoir; The number of flood occurrences and flood losses are obtained according to the historical data of flood defense, and the flood frequency and flood loss rate are obtained, and the flood frequency and the flood loss rate are multiplied to obtain the flood risk coefficient; Obtaining the slope and soil type according to the terrain and geographic data, and classifying the slope and soil type according to the terrain score to obtain a terrain score; calculating the average terrain score of the water conservancy project area according to the terrain score to obtain a terrain vulnerability coefficient; The weighted sum of building density and population density is used to obtain the social vulnerability coefficient; The reservoir storage coefficient, the flood risk coefficient, the terrain vulnerability coefficient and the social vulnerability coefficient are weightedly summed to obtain the vulnerability coefficient.

7. The real-time dispatching method of a water conservancy project system based on neural network technology according to claim 1 is characterized in that: The calculation of the flood control coefficient includes: The rainfall coefficient, the rain cloud influence coefficient and the vulnerability coefficient are normalized respectively, and the normalized rainfall coefficient, the rain cloud influence coefficient and the vulnerability coefficient are weightedly summed to obtain the flood control coefficient.

8. The real-time dispatching method of a water conservancy project system based on neural network technology according to claim 1 is characterized in that: The scheduling constraints include: The maximum and minimum water level constraints for flood control are used to keep the reservoir water level within the maximum and minimum allowable range; Maximum and minimum discharge flow constraints are used to keep reservoir flow within the maximum and minimum allowable ranges; Outflow change constraint, used to keep the reservoir outflow change range within the allowable range; A water balance constraint, used to update the reservoir water storage capacity according to the outflow and inflow; The multi-reservoir coordination constraint is used to coordinate the outflow and inflow of the upstream and downstream reservoirs.

Images